17.
14
Using questionnaires in small scale research
questionnaire to each set of parents. You may
dhcl that both Sandy
Thomson and Sarah Thomson have been picked. In that case you
should send one questionnaire for each, naming the child, because
the Thomsons might have a different view about the advice given
to each child.
In random sampling, each member of the sample is picked from
the whole population, not just from those left after previous
selections. It is possible that the same individual may be picked
twice. In such a case you would not send two questionnaires, but
you would include the data for that individual twice. This may
seem strange, since many individuals are not included at all, but it
is consistent with the statistical laws that decide how well a random
sample represents the population. As you can see, the rules about
sampling, though essentially simple, require to be followed exactly.
WI What size of sample?
We said previously that in choosing the sample randomly your aim
is that all the natural variations within the chosen population will
tend to even out. For that to happen you need fairly large numbers
in the sample. The more varied .your population, the larger the
number needed if you are to be confident about extrapolating from
the sample to the population. Thus, if academic experience and
ability are important factors in your study, then pupils taking
Higher French in a particular school are likely to be more
homogeneous than the schools' inta
of first year pupils, and you
could be content with a smaller sample in the forter case.
Perhaps surprisingly, there are no firm rules abbut sample size.
Most authors suggest 30 as the minimnm, one reason being that
with numbers below 30 the statistica,
formulae may have to be
adapted slightly.
However, with small numbers you would
probably not use a structured questionnaire, and, even if you did,
you would not rely on statistics to strengthen, your conclusions,
perhaps using some interviews to check your interpretation of your
data.
So, we are not suggesting that you should become embroiled with
statistics.
It
is helpful if you have some feeling for what the
numbers involved imply about the degree of confidence with which
you can report results: in other words a grasp of what the statistical
procedures make more precise and quantitative when you use them.
Any randem sample will be an approximate representation of the